🧠 DocMind AI — RAG Document Q&A

An AI-powered document question-answering app built with Retrieval-Augmented Generation (RAG).

Upload any PDF and ask natural language questions — the app finds the most relevant sections and answers using LLaMA 3.3.

🚀 Live Demo

👉 Try it here

🛠️ Tech Stack

• LangChain — RAG pipeline orchestration
• FAISS — Vector database for semantic search
• Groq (LLaMA 3.3 70B) — LLM for answer generation
• HuggingFace Embeddings — Local text embeddings (all-MiniLM-L6-v2)
• Streamlit — Frontend UI
• PyMuPDF — PDF text extraction

⚙️ How It Works

1. Upload a PDF → text is extracted using PyMuPDF
2. Text is split into 300-character chunks with 80-character overlap
3. Chunks are embedded and stored in a FAISS vector index
4. User asks a question → top 6 relevant chunks are retrieved
5. Chunks + question are sent to LLaMA 3.3 via Groq API
6. Answer is displayed with source chunks for transparency

🔧 Run Locally

git clone https://github.com/deekshith-8/rag-doc-qa.git
cd rag-doc-qa
pip install -r requirements.txt

Create a .env file: Run the app:

streamlit run app.py

👤 Built by

Deekshith Gowda — LinkedIn · GitHub